The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes shell commands and manages subprocesses to facilitate testing workflows. The run-agent-verification.ts script uses child_process.spawn and spawnSync to invoke agent engines such as codex or claude and to manage test environment setup. Subcommands like init, update, and ci use npm and npx to install the shiplightai CLI and associated dependencies.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration and tooling from the author's infrastructure and well-known services. It utilizes npx shiplightai@latest and npx skills@latest to maintain the local environment. These resources are provided by the skill's author. CI workflows are configured to use the ci-triage helper from the ShiplightAI GitHub organization. The test runner automates the installation of the Chromium browser via npx playwright install.
[DATA_EXFILTRATION]: The skill transmits diagnostic data and test reports to the Shiplight Cloud platform. The support subcommand collects system metadata and session logs to generate troubleshooting tickets sent to nova-api.shiplight.ai. The ci and cloud subcommands facilitate the upload of test artifacts, including screenshots and execution traces, to the Shiplight Cloud (Nova) backend. The skill includes logic to redact potential secrets from diagnostic data before transmission.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection given its role in processing external data. Ingestion points: Subcommands under references/review/ ingest content from live web pages via browser automation tools. The create-agent-verification subcommand executes instructions defined in Markdown case files. Boundary markers: The skill does not explicitly use delimiters or specialized instructions to ignore potential commands embedded within ingested web content. Capability inventory: The execution environment provides access to shell commands, network operations, and browser interaction tools. Sanitization: The skill implements redaction for support diagnostics but lacks specific sanitization for data ingested during live application reviews.

shiplight